Flexible and accurate inference and learning for deep generative models

TL;DR

This paper presents a novel inference and learning method for hierarchical deep generative models that improves flexibility and accuracy by using an explicit, non-parametric posterior representation and an extended wake-sleep algorithm.

Contribution

It introduces the distributed distributional code Helmholtz machine, a new approach that enhances posterior approximation without restrictive parametric forms, enabling better hierarchical model learning.

Findings

Outperforms state-of-the-art methods on synthetic data

Achieves superior results on natural image patches

Demonstrates effectiveness on MNIST dataset

Abstract

We introduce a new approach to learning in hierarchical latent-variable generative models called the "distributed distributional code Helmholtz machine", which emphasises flexibility and accuracy in the inferential process. In common with the original Helmholtz machine and later variational autoencoder algorithms (but unlike adverserial methods) our approach learns an explicit inference or "recognition" model to approximate the posterior distribution over the latent variables. Unlike in these earlier methods, the posterior representation is not limited to a narrow tractable parameterised form (nor is it represented by samples). To train the generative and recognition models we develop an extended wake-sleep algorithm inspired by the original Helmholtz Machine. This makes it possible to learn hierarchical latent models with both discrete and continuous variables, where an accurate posterior representation is essential. We demonstrate that the new algorithm outperforms current state-of-the-art methods on synthetic, natural image patch and the MNIST data sets.